Micro-magnetic and three-dimensional point cloud fusion-based damage identification method and system for scraper conveyor

Abstract

This invention relates to the field of intelligent detection technology for fully mechanized coal mining equipment, and discloses a method and system for damage identification of scraper conveyors by fusing micromagnetic data and three-dimensional point clouds. The method includes micromagnetic data acquisition, point cloud data acquisition, data preprocessing, feature extraction, data fusion, damage identification, and spatial positioning. This invention determines the no-load inspection status and triggers a self-cleaning device for purging and drying, simultaneously activating the micromagnetic scanning and three-dimensional point cloud acquisition process; it extracts magnetic and geometric spatial anchor points, performs spatial mapping interpolation calculations to generate absolute spatial position coordinates, and establishes an associated mapping matrix to complete coordinate registration and time synchronization; it calculates surface indentation depth parameters to perform inverse threshold compensation calculations on high-frequency damage signals, extracts and splices geometric and magnetic feature vectors to generate a joint feature vector, and inputs the joint feature vector into a target support vector machine classification model to generate damage classification results.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology for coal mine fully mechanized mining equipment, specifically to a method and system for damage identification of scraper conveyors based on the fusion of micromagnetism and three-dimensional point cloud. Background Technology

[0002] During operation, critical components of scraper conveyors experience wear and damage. During damage detection, coal sludge and residual moisture adhering to the component surfaces can cause specular reflection of the structured light projected by the vision sensor, resulting in data interference and reducing the signal-to-noise ratio of multimodal data acquisition.

[0003] When using multiple sensors to acquire multimodal data, the magnetic signal data and 3D point cloud data acquired by different types of sensors deviate in both time and spatial coordinate systems. Due to fluctuations in equipment operating speed and mechanical assembly offsets, it is difficult to achieve accurate spatial coordinate registration and time synchronization of data from different sensors, leading to errors in subsequent data association and fusion.

[0004] Wear and indentation on the surface of key components of scraper conveyors can cause changes in the spatial leakage magnetic field, resulting in baseline drift and distortion of the extracted magnetic signal features. When uncompensated and calibrated feature data is input into the classification model, it is difficult to accurately determine the specific damage type of the component, thus limiting the accuracy of damage identification and spatial positioning in scraper conveyors.

[0005] Therefore, this invention proposes a method and system for damage identification of scraper conveyors that integrates micromagnetism and three-dimensional point cloud to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for damage identification of scraper conveyors by fusing micromagnetic signals and three-dimensional point clouds. This solves the problems of interference from structured light specular reflection caused by residual water film on the surface of key components of scraper conveyors in multimodal data acquisition, inability to achieve coordinate registration and time synchronization between magnetic signal data and three-dimensional point cloud data due to fluctuations in equipment operating speed and mechanical assembly offsets, and errors in damage classification results caused by baseline drift of leakage magnetic field due to surface wear and depressions.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] The first aspect of this invention provides a method for damage identification of scraper conveyors by fusing micromagnetic data with three-dimensional point clouds, including micromagnetic data acquisition, point cloud data acquisition, data preprocessing, feature extraction, data fusion, damage identification, and spatial localization. The specific steps are as follows: The system monitors the real-time operating current value of the scraper conveyor drive motor and compares it with the set no-load current threshold. When the real-time operating current value is less than or equal to the no-load current threshold and the duration reaches the set time threshold, the system determines that the scraper conveyor is in no-load inspection state and sends a linkage trigger command to the self-cleaning device. The self-cleaning device receives the linkage trigger command and sprays high-pressure fluid onto the surface of key components to peel off the adhering substances and expose the metal substrate of the key components. The self-cleaning device then activates the high-pressure air knife to blow and dry the surface of the metal substrate.

[0009] When the system determines that it has the conditions for multimodal high signal-to-noise ratio acquisition, the system simultaneously activates the micro-magnetic scanning and three-dimensional point cloud acquisition process. The system uses a magnetic sensing device based on the giant magnetoresistance effect to acquire magnetic signal data on the surface of key components. The magnetic sensing device arranges multiple giant magnetoresistance sensors at equal intervals along the width direction of the scraper. The giant magnetoresistance sensors use cobalt-based amorphous wires as sensitive elements and continuously acquire magnetic signal data on the surface of key components. The system generates a synchronous pulse trigger signal and transmits the synchronous pulse trigger signal in parallel to the line laser and the high-speed industrial camera. The line laser receives the synchronous pulse trigger signal, generates structured light stripes, and projects them onto the surface of the key components. The high-speed industrial camera receives the synchronous pulse trigger signal, opens the shutter, and captures the structured light stripes to form a laser stripe image.

[0010] The system extracts laser stripe images captured by a high-speed industrial camera. The system performs grayscale conversion on the laser stripe images and uses the grayscale centroid method to extract the center pixel coordinates of the structured light stripes. The system calculates the three-dimensional spatial depth value mapped by the center pixel coordinates based on the principle of triangulation. The system maps the center pixel coordinates and the three-dimensional spatial depth value to the three-dimensional coordinates of the surface points of key components in the world coordinate system. The system aggregates the three-dimensional coordinates of the surface points of key components to generate three-dimensional point cloud data.

[0011] The system acquires magnetic signal data and performs data preprocessing. The system converts the time-domain magnetic signal data into frequency-domain data through Fast Fourier Transform. The system performs low-pass filtering on the frequency-domain data and performs inverse Fast Fourier Transform to generate low-frequency positioning signals. The system performs high-pass filtering on the frequency-domain data and performs inverse Fast Fourier Transform to generate high-frequency damage signals. The system uses wavelet soft thresholding algorithm to denoise the high-frequency damage signals.

[0012] The system reads low-frequency positioning signals and extracts magnetic spatial anchor points. The system reads 3D point cloud data and extracts geometric spatial anchor points. The system extracts timestamp data of the starting and ending magnetic spatial anchor points and calculates the time difference. The system extracts 3D coordinate data of the starting and ending geometric spatial anchor points and calculates the spatial distance. Combining the time difference and spatial distance, the system performs linear velocity interpolation to generate the average speed value of the chain link. The system extracts the mechanical assembly offset parameters between the scraper component and the chain link. Combining the average speed value of the chain link and the mechanical assembly offset parameters, the system performs spatial mapping interpolation to generate the absolute spatial position coordinates. The system establishes a correlation mapping matrix between the high-frequency damage signal and the 3D point cloud data in the same 3D coordinate system.

[0013] The system reads 3D point cloud data and performs point cloud downsampling to generate target point cloud data points. The system calculates the surface normal vector parameters, principal curvature parameters, and Gaussian curvature parameters of the target point cloud data points. It extracts locally flat point clouds that conform to the outer contour features of the key component's reference model as rigid edge point clouds of the unworn area. The system performs spatial registration with the coordinates of the key component's reference model. The system calculates the surface indentation depth parameters and splices the surface normal vector parameters, principal curvature parameters, Gaussian curvature parameters, and surface indentation depth parameters to generate a geometric feature vector. The system performs inverse threshold compensation calculation on the high-frequency damage signal based on the magnetic coupling compensation coefficient and the surface indentation depth parameters to generate a corrected high-frequency signal.

[0014] The system extracts the peak-to-peak value, root mean square value, amplitude gradient parameters, spectral energy parameters, and center frequency parameters of the corrected high-frequency signal to generate magnetic feature vectors. The system extracts aligned coordinate data based on the correlation mapping matrix and matches and merges geometric and magnetic feature vectors with the same absolute spatial coordinates. The system then concatenates the weighted magnetic features and weighted geometric features to generate a joint feature vector.

[0015] The system will input the joint feature vector into the target support vector machine classification model to generate damage classification results. The system will combine the absolute spatial position coordinates to generate damage position mapping records. The system will acquire the sprocket rotation angle data collected by the rotary encoder and generate global spatial three-dimensional coordinates. The system will bind the global spatial three-dimensional coordinates with the damage classification results to generate visualized damage location label data.

[0016] The second aspect of this invention provides a damage identification system for scraper conveyors that integrates micromagnetism and three-dimensional point cloud, including a sensing and acquisition module, a data hub module, a synchronization registration module, a fusion analysis module, and an intelligent output module; The sensing and acquisition module integrates a micro-magnetic sensor array, a line laser, a high-speed industrial camera, a hardware synchronization controller, a rotary encoder, and a zero-point identification sensor; the data hub module receives magnetic signal data and 3D point cloud data and extracts synchronization pulse trigger signals to generate initial timestamp labels; the synchronization registration module parses the initial timestamp labels, performs time axis alignment operations, and generates a coarsely aligned multimodal dataset; the fusion analysis module extracts geometric feature vectors and magnetic feature vectors, performs splicing operations to generate joint feature vectors, and calls the target support vector machine classification model to generate damage classification results; the intelligent output module binds the damage classification results and global spatial 3D coordinates to generate visualized damage location label data.

[0017] This invention provides a method and system for damage identification of scraper conveyors based on the fusion of micromagnetic fields and three-dimensional point clouds. It has the following beneficial effects: 1. This invention determines the no-load inspection status by monitoring the real-time operating current value of the scraper conveyor drive motor, and links the self-cleaning device to spray high-pressure fluid to peel off the attached substances, and links the activation of the high-pressure air knife to blow and dry the surface of the metal substrate, thereby eliminating the interference of structured light specular reflection caused by the residual water film on the surface, and ensuring high signal-to-noise ratio acquisition conditions for magnetic signal data and three-dimensional point cloud data.

[0018] 2. This invention extracts the magnetic space anchor points corresponding to low-frequency positioning signals and the geometric space anchor points corresponding to three-dimensional point cloud data, and performs spatial mapping interpolation calculations to generate absolute spatial position coordinates by combining the average speed value of chain link operation and mechanical assembly offset parameters. It establishes an association mapping matrix between high-frequency damage signals and three-dimensional point cloud data in the same three-dimensional coordinate system, and completes the coordinate registration and time synchronization of magnetic signal data and three-dimensional point cloud data.

[0019] 3. This invention generates surface depression depth parameters by calculating the vertical coordinate difference between the target point cloud data points and the reference model coordinates of key components. Based on the magnetic coupling compensation coefficient and the surface depression depth parameters, it performs inverse threshold compensation calculation on high-frequency damage signals, eliminating the baseline drift phenomenon of leakage magnetic field caused by surface wear depression. Furthermore, it concatenates and splices geometric feature vectors and magnetic feature vectors to generate joint feature vectors, which are then input into the target support vector machine classification model to output specific damage classification results. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the present invention; Figure 3 This is a comparison diagram of the magnetic coupling compensation effect of high-frequency damage signals in a specific application embodiment of the present invention. Detailed Implementation

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] See attached document Figure 1 This invention provides a method for damage identification of scraper conveyors based on the fusion of micromagnetic data and 3D point clouds. This method is applied to a scraper conveyor damage identification system based on the fusion of micromagnetic data and 3D point clouds, and includes micromagnetic data acquisition, point cloud data acquisition, data preprocessing, feature extraction, data fusion, damage identification, and spatial localization. The specific steps are as follows: S1. The system monitors the real-time operating current value of the scraper conveyor drive motor and compares it with the set no-load current threshold. When the real-time operating current value is less than or equal to the no-load current threshold and the duration reaches the set time threshold, the system determines that the scraper conveyor is in no-load inspection state and sends a linkage trigger command to the self-cleaning device. The self-cleaning device receives the linkage trigger command and sprays high-pressure fluid onto the surface of key components to peel off the adhering substances and expose the metal substrate of the key components. Subsequently, the self-cleaning device activates a high-pressure air knife to blow and dry the surface of the metal substrate to eliminate the structured light specular reflection interference caused by the residual water film on the surface.

[0023] ; In the formula, This indicates the trigger flag for no-load inspection status. Represents the real-time operating current value; Represents the no-load current threshold; This represents the duration during which the real-time operating current value is less than or equal to the no-load current threshold. This represents the time threshold.

[0024] S2, when When the value equals 1, the system determines that it has the conditions for multimodal high signal-to-noise ratio acquisition. The system reads the spatial distance parameter between the high-pressure air knife and the observation field of the line laser, and calculates the current running speed of the scraper conveyor by combining the pulse frequency output by the rotary encoder in real time. The system calculates the surface purging safety waiting time and delays the surface purging safety waiting time after the self-cleaning device is activated. The system simultaneously activates the micro-magnetic scanning and three-dimensional point cloud acquisition process. The system uses a magnetic sensing device based on the giant magnetoresistance effect to collect magnetic signal data on the surface of key components. The magnetic sensing device arranges multiple giant magnetoresistance sensors at equal intervals along the width direction of the scraper. The giant magnetoresistance sensors use cobalt-based amorphous wire as the sensitive element and set the excitation frequency to continuously collect magnetic signal data on the surface of key components from 1MHz to 10MHz.

[0025] The system drives the magnetic sensing device to input an AC excitation current into the cobalt-based amorphous wire of the giant magnetoresistive sensor and generate an alternating magnetic field in the cobalt-based amorphous wire. The alternating magnetic field produces a skin effect on the surface of the cobalt-based amorphous wire and forms a skin depth. Defects inside the key components generate a spatial leakage magnetic field. The spatial leakage magnetic field changes the permeability of the cobalt-based amorphous wire, and the change in permeability causes a change in the skin depth.

[0026] ; In the formula, Represents skin depth; Represents the resistivity of cobalt-based amorphous wire; Represents pi; The frequency of the alternating current excitation current; This represents the permeability of a cobalt-based amorphous filament.

[0027] The change in skin depth causes a change in the impedance value of the giant magnetoresistive sensor. The system extracts the impedance change and converts it into a voltage signal. The system performs analog-to-digital conversion on the voltage signal and samples and stores it as magnetic signal data.

[0028] The system generates a synchronous pulse trigger signal, which is then transmitted in parallel to the line laser and the high-speed industrial camera via hardware cables.

[0029] A line laser receives a synchronous pulse trigger signal to generate structured light stripes and projects them onto the surface of a key component. The line laser projects structured light stripes with a wavelength of 650nm and a power of 10mW to 50mW. A high-speed industrial camera receives the synchronous pulse trigger signal, opens the shutter to expose and capture the structured light stripes to form a laser stripe image. The high-speed industrial camera captures the structured light stripes with a frame rate of 1000fps to 5000fps.

[0030] ; In the formula, This represents the time difference between synchronous triggering; This represents the shutter opening time of a high-speed industrial camera. The time points at which the representative line laser generates structured light fringes.

[0031] S3. The system extracts the laser stripe image captured by the high-speed industrial camera. The system performs grayscale conversion on the laser stripe image and uses the grayscale centroid method to extract the center pixel coordinates of the structured light stripe. The system calculates the three-dimensional spatial depth value mapped by the center pixel coordinates based on the triangulation principle.

[0032] ; In the formula, Represents the depth value in three-dimensional space; The baseline distance between the line laser and the high-speed industrial camera; The focal length represents that of a high-speed industrial camera. Represents the center pixel coordinates; Represents the reference coordinates of the image center of a high-speed industrial camera.

[0033] The system acquires 3D spatial depth values ​​and, based on the pre-calibrated intrinsic and extrinsic parameter matrix of a high-speed industrial camera, maps the center pixel coordinates and 3D spatial depth values ​​to 3D coordinates of key component surface points in the world coordinate system. The system then aggregates these 3D coordinates to generate 3D point cloud data.

[0034] S4. The system acquires magnetic signal data and performs data preprocessing. The system converts the time-domain magnetic signal data into frequency-domain data using a fast Fourier transform. The system uses a low-pass filter and sets a low-pass cutoff frequency to perform low-pass filtering on the frequency-domain data. The system performs an inverse fast Fourier transform on the low-pass filtered frequency-domain data to generate a low-frequency positioning signal. The system uses a high-pass filter and sets a high-pass cutoff frequency to perform high-pass filtering on the frequency-domain data. The system performs an inverse fast Fourier transform on the high-pass filtered frequency-domain data to generate a high-frequency damage signal.

[0035] The system employs a wavelet soft thresholding algorithm to denoise high-frequency damage signals. The system selects wavelet basis functions and sets the number of decomposition levels to perform wavelet decomposition on the high-frequency damage signals to extract approximation coefficients and detail coefficients. The system uses a soft thresholding function to shrink the detail coefficients according to the set fixed threshold to generate target detail coefficients. The system merges the approximation coefficients and target detail coefficients and performs inverse wavelet transform reconstruction to generate the final high-frequency damage signal.

[0036] S5. The system generates a data time hysteresis compensation amount by performing a division operation based on the installation spacing parameters of the micro-magnetic sensor array and the line laser in the running direction of the scraper conveyor, combined with the average speed value of the chain links. It then performs offset compensation on the timestamps of the magnetic signal data to achieve initial macroscopic coarse alignment. Subsequently, the system reads the low-frequency positioning signal and performs first-order difference calculation according to the time series to obtain the signal gradient value sequence. The system identifies the zero-crossing points in the signal gradient value sequence where the values ​​change from positive to negative, extracts the timestamp data corresponding to the zero-crossing points, and records them as magnetic spatial anchor points. The system sequentially extracts the magnetic spatial anchor points corresponding to two adjacent scraper components along the time series, defining the earlier magnetic spatial anchor point in the time series as the starting magnetic spatial anchor point and the later magnetic anchor point in the time series as the ending magnetic spatial anchor point.

[0037] Because the metal volume of the scraper component of the scraper conveyor is much larger than that of ordinary chain links, when the scraper passes through the micro magnetic sensor array, its large volume of ferromagnetic material causes a violent alternation of the macroscopic spatial leakage magnetic field; at this time, the rate of change of magnetic flux reaches an extreme value, which in a sense represents the geometric symmetry plane of the scraper passing orthogonally through the cross section of the sensor array, thereby establishing an equivalent mapping relationship between the low-frequency micro magnetic zero-crossing point and the mechanical center of the scraper.

[0038] The system reads 3D point cloud data and performs height difference calculations in the running direction of the scraper conveyor to generate a height gradient matrix. The system compares the elements in the height gradient matrix with a set gradient threshold and extracts abrupt coordinate data points that exceed the threshold. The system then uses these abrupt coordinate data points to fit and generate a scraper component contour model. The system calculates the center 3D coordinates of the scraper component contour model and records them as geometric space anchor points. The system extracts geometric space anchor points that match the two adjacent scraper components, defining the geometric space anchor point in the forward running direction as the starting geometric space anchor point and the geometric space anchor point in the backward running direction as the ending geometric space anchor point.

[0039] The system extracts the timestamp data of the starting and ending magnetic space anchor points and calculates the time difference. The system extracts the three-dimensional coordinate data of the starting and ending geometric space anchor points and calculates the spatial distance. The system extracts the high-frequency pulse subdivision data output by the rotary encoder, and performs linear velocity interpolation calculation by combining the spatial distance value and the time difference value to generate the average speed value of the chain link. The system reads the sampling timestamp of the high-frequency damage signal. The system extracts the mechanical assembly offset parameters between the scraper component and the chain link. The system combines the average speed value of the chain link and the mechanical assembly offset parameters to perform spatial mapping interpolation calculation to generate the absolute spatial position coordinates.

[0040] ; In the formula, Represents absolute spatial position coordinates; Three-dimensional coordinate data representing the initial geometrical anchor point; Represents spatial distance value; Represents the time difference; The sampling timestamp represents the high-frequency damage signal; Timestamp data representing the initial magnetic space anchor point; This represents the mechanical assembly offset parameter.

[0041] Among them, mechanical assembly offset parameters Specifically, this is characterized by the fixed installation error between the symmetry center plane of the scraper and the scanning section of the micro-magnetic sensor array, as well as the static tolerance of the mechanical assembly between the chain link and the scraper.

[0042] The system inputs the absolute spatial position coordinates into the coordinate system of the three-dimensional point cloud data for overlap matching calculation. The system establishes an association mapping matrix between the high-frequency damage signal and the three-dimensional point cloud data in the same three-dimensional coordinate system. The system outputs the association mapping matrix to complete the coordinate registration and time synchronization of the magnetic signal data and the three-dimensional point cloud data.

[0043] S6. The system reads 3D point cloud data and performs point cloud downsampling to generate target point cloud data points. The system uses principal component analysis to calculate the surface normal vector parameters of the target point cloud data points. Based on the surface normal vector parameters, the system calculates the principal curvature parameters and Gaussian curvature parameters of the target point cloud data points, and uses the calculated surface normal vector parameters, principal curvature parameters, and Gaussian curvature parameters as surface parameters.

[0044] The system reads the coordinates of the key component's reference model. It sets a curvature change rate threshold and a flatness threshold, and filters the target point cloud data points based on surface normal vector parameters and Gaussian curvature parameters. Locally flat point clouds that conform to the outer contour characteristics of the key component's reference model are extracted as rigid edge point clouds of the unworn area. The rigid edge point clouds are then spatially registered with the key component's reference model coordinates. The system uses an iterative nearest-point algorithm or a normal distribution transformation algorithm to perform the spatial registration operation, eliminating translational and rotational spatial deviations caused by the vibration of the scraper conveyor. After registration, the system compares the 3D coordinates of the target point cloud data points with the coordinates of the key component's reference model to calculate the vertical coordinate difference and generate surface indentation depth parameters. The system extracts surface parameters from the registered 3D point cloud data and stitches the surface parameters and surface indentation depth parameters to generate a geometric feature vector.

[0045] The system reads the lateral arrangement coordinates and spacing parameters of multiple giant magnetoresistive sensors in the micro-magnetic sensor array along the width of the scraper. Based on the lateral arrangement coordinates and spacing parameters, the system divides the coordinate system of the 3D point cloud data into multiple mapping data bands along the width of the scraper. The system extracts the corresponding surface indentation depth parameters in each mapping data band. Based on the lateral position index of the mapping data band, the system extracts the associated high-frequency damage signal of the corresponding channel in the micro-magnetic sensor array. Surface wear of key components causes changes in cross-sectional area and generates a leakage magnetic field baseline drift phenomenon. The system accesses the material calibration database and reads the magnetic shape coupling compensation coefficient corresponding to the material of the key component. The magnetic shape coupling compensation coefficient is obtained by the system scanning standard parts of the same material with known wear depths in advance and fitting the mapping relationship curve between the leakage magnetic field baseline drift and the wear depth. Based on the magnetic shape coupling compensation coefficient and the surface indentation depth parameters, the system performs inverse threshold compensation calculation on the high-frequency damage signal to generate a corrected high-frequency signal.

[0046] ; In the formula, This represents a correction of the high-frequency signal; Represents high-frequency damage signals; Represents the magnetic coupling compensation coefficient; This represents the depth parameter of the surface depression.

[0047] Among them, the magnetic coupling compensation coefficient The sign depends on whether the output characteristics and lift-off effect of the giant magnetoresistive sensor are positively or negatively correlated. Its engineering significance lies in compensating for the baseline offset of the leakage magnetic signal caused by surface depression to the reference level.

[0048] S7. The system reads the corrected high-frequency signal and sets the corresponding time-domain sliding window based on the frame rate period of the laser stripe image captured by the high-speed industrial camera. The system extracts the peak-to-peak value and root mean square value of the signal in each time-domain sliding window. The system performs first-order difference calculation based on the time series of the corrected high-frequency signal to generate amplitude abrupt change gradient parameters.

[0049] The system performs a discrete Fourier transform on the corrected high-frequency signal to generate a frequency domain data sequence. Based on the frequency domain data sequence, the system calculates the frequency energy parameters and center frequency parameters after the frequency domain transformation within a set frequency band. The system extracts the signal peak-to-peak value, signal root mean square value, amplitude gradient parameters, and the frequency energy parameters and center frequency parameters after the frequency domain transformation. The system then concatenates the signal peak-to-peak value, signal root mean square value, amplitude gradient parameters, frequency energy parameters, and center frequency parameters to generate a magnetic feature vector.

[0050] The system extracts geometric and magnetic feature vectors, reads the correlation mapping matrix to extract alignment coordinate data, and matches and merges geometric and magnetic feature vectors with the same absolute spatial coordinates based on the alignment coordinate data.

[0051] The system sets the magnetic signal weight coefficient and the point cloud weight coefficient. The system calculates the product of the magnetic feature vector and the magnetic signal weight coefficient to generate a weighted magnetic feature. The system calculates the product of the geometric feature vector and the point cloud weight coefficient to generate a weighted geometric feature. The system concatenates and splices the weighted magnetic feature and the weighted geometric feature to generate a joint feature vector.

[0052] ; In the formula, Represents the joint eigenvector; Represents the magnetic signal weighting coefficient; Represents the magnetic characteristic vector; Represents the point cloud weight coefficient; Represents the geometric characteristic vector.

[0053] S8. The system extracts the joint feature vector and constructs a pre-set damage classification model, specifically a support vector machine (SVM) classification model. The system acquires sample data containing historical damage features and corresponding label data. The label data includes at least classification labels such as normal wear, matrix microcracks, plastic deformation, and chain breakage edges. The system inputs the sample data and label data into the SVM classification model to perform iterative training and generate a target SVM classification model. The system inputs the joint feature vector into the target SVM classification model. The target SVM classification model calls a kernel function to map the joint feature vector to a high-dimensional feature space and calculates the classification boundary bias value. The system generates the damage classification result based on the classification boundary bias value.

[0054] ; In the formula, This represents the damage classification result; Represents a symbolic function; This represents the number of support vectors; Represents the Lagrange multiplier; Represents tag data; Represents the kernel function; Represents the joint eigenvector; Represents support vectors; This represents the classification boundary bias value.

[0055] The system reads the damage classification results to determine the damage category of key components. The system combines the absolute spatial coordinates to generate a damage location mapping record. The system stores the damage location mapping record and the damage classification results in the database to complete damage identification and spatial positioning.

[0056] The system acquires sprocket rotation angle data collected by the rotary encoder. Based on the sprocket rotation angle data, the system calculates the cumulative number of sprocket rotations. The system reads the cumulative length of the polygonal pitch of a single sprocket rotation and the encoder's initial reference coordinates. The system extracts the absolute spatial position coordinates mapped to the damaged area. Based on the encoder's initial reference coordinates, the cumulative number of sprocket rotations, the cumulative length of the polygonal pitch of a single sprocket rotation, and the absolute spatial position coordinates, the system performs a linear superposition operation to generate global three-dimensional coordinates.

[0057] ; In the formula, Represents the three-dimensional coordinates of the global space; Represents the initial reference coordinate parameters of the encoder; This parameter represents the cumulative number of rotations of the sprocket. The parameter representing the cumulative length of the polygonal pitch of a single turn of the sprocket; Represents absolute spatial coordinates.

[0058] The system places an absolute position mark on the reference scraper of the scraper chain and sets up an associated zero-point identification sensor. When the zero-point identification sensor detects that the absolute position mark has passed, the system issues a zeroing command to record the cumulative number of rotations of the sprocket. Zeroing eliminates the cumulative error in global three-dimensional coordinates caused by the flexible stretching and mechanical wear of the scraper chain.

[0059] The system binds global spatial 3D coordinates with damage classification results to generate visualized damage location label data. The system then stores the visualized damage location label data in the database to complete the spatial location operation of the damage site.

[0060] See attached document Figure 2 This invention also provides a damage identification system for scraper conveyors that integrates micromagnetic sensors and three-dimensional point clouds, used to perform a method for damage identification of scraper conveyors by integrating micromagnetic sensors and three-dimensional point clouds. The system specifically includes: a sensing and acquisition module, a data hub module, a synchronization registration module, a fusion analysis module, and an intelligent output module.

[0061] The sensing and acquisition module integrates a micro magnetic sensor array, a line laser, a high-speed industrial camera, a hardware synchronization controller, a rotary encoder, and a zero-point identification sensor.

[0062] A micro-magnetic sensor array is installed at the base of the sprocket at the head of the scraper conveyor. The array is connected to the system host via shielded cables and transmits magnetic signal data. A line laser and a high-speed industrial camera are fixed to a protective bracket directly above the scraper conveyor head, forming a predetermined installation angle. A linear polarizing filter is installed at the front of the high-speed industrial camera lens to suppress high-light diffuse reflection from metal surfaces and residual water droplets, ensuring the integrity of the grayscale features in the laser stripe image.

[0063] The hardware synchronization controller is connected to the micro magnetic sensor array, the line laser, and the high-speed industrial camera respectively through a data communication interface. The hardware synchronization controller transmits synchronization pulse trigger signals in parallel to the micro magnetic sensor array, the line laser, and the high-speed industrial camera.

[0064] The rotary encoder is coaxially mounted on the output shaft of the drive motor at the head of the scraper conveyor. The zero-point identification sensor is fixedly mounted on the reference position of the side wall of the scraper conveyor head or the sprocket cover. The zero-point identification sensor is connected to the system host through a data communication interface to capture the absolute position mark on the scraper chain in real time.

[0065] The system consists of a data hub module, a synchronization and registration module, and a cache database. The data hub module connects to the sensing and acquisition module via a data bus and receives magnetic signal data and 3D point cloud data. The data hub module extracts the synchronization pulse trigger signal generated by the hardware synchronization controller and generates an initial timestamp tag. The data hub module stores the magnetic signal data, 3D point cloud data, and the initial timestamp tag in the cache database.

[0066] The synchronization registration module accesses the cached database to read magnetic signal data and 3D point cloud data. It parses the initial timestamp labels and performs time axis alignment calculations. The system reads the installation spacing parameters of the micro-magnetic sensor array and the line laser in the running direction of the scraper conveyor, and performs a division operation based on the average speed value of the chain links to generate a data time hysteresis compensation amount. The synchronization registration module performs a sliding window offset calculation on the initial timestamp labels of the magnetic signal data according to the data time hysteresis compensation amount to generate aligned timestamp labels. The synchronization registration module filters data streams whose aligned timestamp labels match the initial timestamp labels of the 3D point cloud data, and performs data fusion and packaging operations to generate a coarsely aligned multimodal dataset with macroscopic time hysteresis compensation. The synchronization registration module outputs the coarsely aligned multimodal dataset for subsequent calculations, allowing the system to further extract magnetic and geometric spatial anchor points and perform interpolation calculations to complete the precise spatial registration task.

[0067] The system consists of a fusion analysis module, an intelligent output module, and a human-computer interaction terminal. The fusion analysis module connects to the synchronous registration module and receives synchronous multimodal datasets. The fusion analysis module extracts geometric and magnetic feature vectors from the synchronous multimodal datasets. The fusion analysis module performs splicing operations to generate joint feature vectors. The fusion analysis module calls a pre-set damage classification model to perform operations on the joint feature vectors to generate damage classification results.

[0068] The intelligent output module receives damage classification results and global spatial 3D coordinates. It binds the damage classification results and global spatial 3D coordinates to generate visualized damage location label data. The intelligent output module stores the visualized damage location label data and damage classification results in the database. The intelligent output module transmits the visualized damage location label data to the human-computer interaction terminal. The human-computer interaction terminal renders the 3D model of the scraper conveyor and overlays the visualized damage location label data at the corresponding coordinate positions to complete the operation process of the intelligent output architecture.

[0069] The present invention provides an electronic device, which has a processor, a memory, and a communication interface. The processor, the memory, and the communication interface are interconnected through a data bus to complete instruction transmission and data interaction.

[0070] The memory stores computer program code. The processor reads the computer program code stored in the memory and executes the operation logic set in the code. The processor completes the data processing operation according to the operation logic to realize the damage identification method of scraper conveyor based on the fusion of micromagnetism and three-dimensional point cloud.

[0071] The communication interface receives measurement data sent by the sensing and acquisition module and transmits the measurement data to the processor. The processor performs feature extraction and fusion operations on the measurement data to generate visual damage location tag data. The processor sends the visual damage location tag data to the human-computer interaction terminal for interface display through the communication interface.

[0072] This invention provides a computer-readable storage medium that stores computer program instructions. A processor reads and executes the computer program instructions to complete the computational steps of the scraper conveyor damage identification method based on the fusion of micromagnetism and three-dimensional point cloud.

[0073] Computer-readable storage media types are classified into read-only memory, random access memory, magnetic disks, and optical disks. Computer program instructions are compiled from software code and written into the internal storage array of the computer-readable storage media. Hardware devices read the software code in the computer-readable storage media and convert it into machine-recognizable low-level logic level signals. The low-level logic level signals drive the processor to schedule hardware resources to perform data fusion, feature extraction, and state classification operations.

[0074] Specific application examples: To further aid in understanding the specific implementation logic and working principle of the present invention, the following provides a specific application example and experimental verification effect based on a daily no-load inspection scenario of a scraper conveyor in a fully mechanized coal mining face.

[0075] In this application scenario, it is necessary to evaluate the wear and internal microcrack damage of key components of the scraper conveyor (such as scraper chains and sprockets) under complex working conditions, and at the same time verify the system's ability to correct for baseline drift of the micro-magnetic leakage field. To illustrate the data calculation and condition assessment process within the system, the following section will use the physical test parameters from the scenario to perform calculations in the data hub module and the fusion analysis module.

[0076] First, the system monitors the real-time operating current value of the scraper conveyor drive motor. The set no-load current threshold is known. Equal to 50A, the set time threshold Equals 20 seconds. During the inspection period, the system extracts the current real-time operating current value. The value is 42A, and the duration of this state is... It reached 25 seconds. Due to... and The logical condition is met: .

[0077] The system determines the trigger flag for no-load inspection status. If the value equals 1, it indicates that the scraper conveyor is in an unloaded inspection state and triggers the self-cleaning device and high-pressure air knife.

[0078] The next stage involves 3D point cloud data acquisition and processing. The baseline distance between the line laser and the high-speed industrial camera is also considered. The focal length of a high-speed industrial camera is 200mm. The length is 16mm. The system extracts the center pixel coordinates of a structured light fringe. Given the coordinates of the image center of a high-speed industrial camera (value 800),... The value is 500. The system calculates the three-dimensional spatial depth value mapped to the coordinates of the center pixel based on the principle of triangulation. : ; This three-dimensional spatial depth value is further converted into three-dimensional coordinates of key component surface points, which are used to extract surface depression depth parameters.

[0079] During the data fusion and coordinate registration stage, the system needs to calculate the absolute spatial coordinates. Given the three-dimensional coordinates of the initial geometrical anchor point. The corresponding one-dimensional running coordinate is 1000mm, the spatial distance between the starting geometric space anchor point and the ending geometric space anchor point. The time difference between the initial and final magnetic space anchor points is 500mm. The sampling timestamp is 0.5s. This is the time stamp of a high-frequency damage signal extracted. The time stamp data corresponding to the starting magnetic space anchor point is 1.2s. The time is 1.0s, and the mechanical assembly offset parameter is... The calibration value is 5mm. The average speed value of the system combined with the chain link operation ( Perform spatial mapping interpolation calculations: ; Due to surface wear of key components causing changes in cross-sectional area and resulting in baseline drift of the leakage magnetic field, the system extracts the surface indentation depth parameter at this absolute spatial location from the 3D point cloud. (Note: here) The depth of the depression (referring to the depth of the indentation) is 2.5 mm. The system reads the magnetic coupling compensation coefficient obtained from the material calibration database. The value is 10mV / mm. The system extracts the uncompensated high-frequency damage signal at this location. The value is 150mV, and a reverse threshold compensation calculation is performed to generate a corrected high-frequency signal. : ; Finally, spatial positioning is performed, based on the known initial reference coordinate parameters of the encoder. The parameter is 0mm, representing the cumulative number of rotations of the sprocket. The cumulative length parameter of the polygonal pitch of a single sprocket rotation is 50 revolutions. The value is 2000mm. The system is based on the previously calculated absolute spatial position coordinates. (1205mm), generate global three-dimensional coordinates One-dimensional expansion value: ; The system binds the coordinates to the damage classification results generated by the support vector machine model, thus completing the storage of the visualized damage location label data.

[0080] To demonstrate the effectiveness of the magnetic coupling compensation algorithm of the present invention, a comparative verification scenario was constructed using an experimental platform. This scenario simulated a scraper metal specimen with gradual surface concave wear and internal microcrack damage. The corrected high-frequency signal of the present invention was compared with the original uncompensated high-frequency damage signal, and corresponding operating condition charts were output.

[0081] The conclusions are as follows: Reference Appendix Figure 3 , attached Figure 3 The solid line represents the original high-frequency damage signal (without point cloud concavity parameter compensation), and the dashed line represents the corrected high-frequency signal after processing by the fusion system of this invention.

[0082] From the appendix Figure 3 As can be seen, the original high-frequency damage signal, within the spatial coordinate range of 20mm to 80mm, exhibits a significant leakage magnetic field baseline drift (displaying a raised, broad peak) due to the increased depth of the surface depression. This macroscopic morphological interference easily obscures the true microcrack damage features (sharp peaks) at 50mm, leading to misjudgments in subsequent feature extraction. However, the system of this invention accurately extracts the surface depression depth parameters using 3D point cloud data and performs inverse threshold compensation by incorporating the magnetic shape coupling compensation coefficient. The resulting corrected high-frequency signal eliminates the baseline drift caused by wear depressions, allowing microcrack features to be clearly exposed on the horizontal baseline. This invention's micromagnetic and 3D point cloud fusion system decouples morphological interference from material damage, improving the signal-to-noise ratio and accuracy of damage identification.

Claims

1. A method for damage identification of an apron conveyor by fusing micro-magnetic and three-dimensional point cloud, characterized in that, Includes the following steps: Acquire magnetic signal data and 3D point cloud data of key components of the scraper conveyor; The magnetic signal data is preprocessed to generate low-frequency positioning signals and high-frequency damage signals. Magnetic spatial anchors and geometric spatial anchors are generated based on the low-frequency positioning signals and the three-dimensional point cloud data, respectively. Calculate the absolute spatial position coordinates based on the magnetic space anchor point and the geometric space anchor point, and establish the correlation mapping matrix between the high-frequency damage signal and the three-dimensional point cloud data; The three-dimensional point cloud data is spatially registered with the coordinates of the preset key component reference model and the vertical coordinate difference is calculated to generate the surface depression depth parameter. The geometric feature vector is generated by combining the surface parameters extracted from the registered three-dimensional point cloud data. Based on the surface indentation depth parameter, the high-frequency damage signal is subjected to inverse threshold compensation to generate a corrected high-frequency signal, and the features of the corrected high-frequency signal are extracted to generate a magnetic feature vector; Based on the correlation mapping matrix, the geometric feature vectors with the same absolute spatial coordinates are merged with the magnetic feature vectors to generate a joint feature vector, which is then input into a pre-set damage classification model to generate damage classification results.

2. The micro-magnetic and three-dimensional point cloud fusion-based scraper conveyor damage identification method of claim 1, wherein, The steps for acquiring magnetic signal data and three-dimensional point cloud data of key components of the scraper conveyor include: If the scraper conveyor is determined to be in an unloaded inspection state, the real-time operating current value of the drive motor is monitored and compared with the set unloaded current threshold. When the real-time operating current value is less than or equal to the unloaded current threshold and the duration reaches the set time threshold, the self-cleaning device is controlled to spray high-pressure fluid onto the surface of key components and the high-pressure air knife is turned on to blow and dry. A synchronous pulse trigger signal is generated and transmitted in parallel to a line laser and a high-speed industrial camera. The line laser is controlled to project structured light stripes, and the high-speed industrial camera is controlled to capture and generate the three-dimensional coordinates of key component surface points using the principle of triangulation. The three-dimensional point cloud data is then generated. The driving magnetic sensing device inputs an AC excitation current into the cobalt-based amorphous wire of the giant magnetoresistive sensor to generate an alternating magnetic field, extracts the impedance value change and converts it into a voltage signal to sample and generate the magnetic signal data.

3. The micro-magnetic and three-dimensional point cloud fusion-based scraper conveyor damage identification method of claim 1, wherein, The steps of preprocessing the magnetic signal data to generate low-frequency positioning signals and high-frequency damage signals include: The magnetic signal data in the time domain is converted into frequency domain data using a fast Fourier transform. The frequency domain data is subjected to low-pass filtering and inverse fast Fourier transform is then performed to generate the low-frequency positioning signal. The high-frequency domain data is processed by high-pass filtering and inverse fast Fourier transform is performed to generate the high-frequency damage signal. The approximation coefficients and detail coefficients are extracted by wavelet soft thresholding algorithm to denoise the high-frequency damage signal.

4. The micro-magnetic and three-dimensional point cloud fusion based scraper conveyor damage identification method of claim 1, wherein, The steps of generating magnetic space anchor points and geometric space anchor points based on the low-frequency positioning signal and the three-dimensional point cloud data respectively include: The low-frequency positioning signal is subjected to first-order difference calculation according to the time series to obtain the signal gradient value sequence. The zero-crossing position where the value changes from positive to negative in the signal gradient value sequence is identified, and the timestamp data corresponding to the zero-crossing position is extracted and recorded as the magnetic space anchor point. In the direction of operation, the height difference calculation of the three-dimensional point cloud data is performed to generate a height gradient matrix. The abrupt coordinate data points that exceed the set gradient threshold are extracted and fitted together to generate the outline model of the scraper component. The center three-dimensional coordinates of the outline model of the scraper component are calculated and recorded as the geometric space anchor point.

5. The micro-magnetic and three-dimensional point cloud fusion based scraper conveyor damage identification method of claim 1, wherein, The steps for establishing the correlation mapping matrix between the high-frequency damage signal and the three-dimensional point cloud data include: Extract the timestamp data of adjacent starting magnetic space anchor points and ending magnetic space anchor points to calculate the time difference value; extract the three-dimensional coordinate data of adjacent starting geometric space anchor points and ending geometric space anchor points to calculate the spatial distance value. The average speed of the chain link is generated by linear velocity interpolation based on the spatial distance value and the time difference value. The sampling timestamp of the high-frequency damage signal is read and the mechanical assembly offset parameter is extracted. The spatial mapping interpolation calculation is performed by combining the average speed value of the chain link and the mechanical assembly offset parameter to generate the absolute spatial position coordinates. The absolute spatial position coordinates are input into the coordinate system of the three-dimensional point cloud data for overlap matching calculation to establish the association mapping matrix containing the alignment coordinate data.

6. The micro-magnetic and three-dimensional point cloud fusion based flight bar conveyor damage identification method of claim 1, wherein, The step of generating a geometric feature vector by combining surface parameters extracted from the registered 3D point cloud data includes: Extract locally flat point clouds from the three-dimensional point cloud data that conform to the outer contour features of the reference model of the key component as rigid edge point clouds of the unworn area, and perform spatial registration between the rigid edge point clouds and the coordinates of the reference model of the key component. Perform point cloud downsampling on the registered 3D point cloud data to generate target point cloud data points, and calculate the surface normal vector parameters, principal curvature parameters and Gaussian curvature parameters of the target point cloud data points as the surface parameters; By comparing the three-dimensional coordinates of the target point cloud data points with the registered coordinates of the key component reference model, the vertical coordinate difference is calculated to generate the surface indentation depth parameter. The surface parameter and the surface indentation depth parameter are then combined to generate the geometric feature vector.

7. The micro-magnetic and three-dimensional point cloud fusion based scraper conveyor damage identification method of claim 1, wherein, The step of generating a corrected high-frequency signal by performing inverse threshold compensation on the high-frequency damage signal based on the surface indentation depth parameter includes: The coordinate system of the three-dimensional point cloud data is divided into multiple mapping data zones based on the lateral arrangement coordinates and spacing parameters of the micro-magnetic sensor array; The corresponding surface depression depth parameters are extracted in each of the mapping data bands, and the associated high-frequency damage signals of the corresponding channels in the micro magnetic sensor array are extracted according to the lateral position index of the mapping data bands. Read the magnetic coupling compensation coefficient corresponding to the material of the key component, and perform reverse threshold compensation calculation on the associated high-frequency damage signal based on the magnetic coupling compensation coefficient and the surface indentation depth parameter to generate the corrected high-frequency signal.

8. The micro-magnetic and three-dimensional point cloud fusion based scraper conveyor damage identification method of claim 1, wherein, The step of merging the geometric feature vectors with the same absolute spatial coordinates and the magnetic feature vectors to generate a joint feature vector includes: Extract the peak-to-peak value, root mean square value, amplitude gradient parameters, frequency energy parameters and center frequency parameters of the corrected high-frequency signal, and concatenate them to generate the magnetic feature vector. Based on the alignment coordinate data of the association mapping matrix, the geometric feature vectors and magnetic feature vectors with the same absolute spatial position coordinates are matched and combined; Weighted magnetic features and weighted geometric features are generated by assigning weight coefficients to the magnetic feature vector and the geometric feature vector respectively, and the weighted magnetic features and the weighted geometric features are concatenated and spliced ​​to generate the joint feature vector.

9. The micro-magnetic and three-dimensional point cloud fusion based flight bar conveyor damage identification method of claim 1, wherein, The pre-set damage classification model is a target support vector machine classification model; After the pre-set damage classification model generates the damage classification result, the following is also included: Obtain the parameters of the cumulative number of rotations of the sprocket, the cumulative length of the polygonal pitch of a single rotation of the sprocket, and the initial reference coordinates of the encoder; A global three-dimensional spatial coordinate is generated by performing a linear superposition operation based on the encoder's initial reference coordinate parameters, the sprocket's cumulative rotation number parameters, the sprocket's single-rotation polygonal pitch cumulative length parameters, and the absolute spatial position coordinates. The global spatial three-dimensional coordinates are data-bound with the damage classification results to generate visualized damage location label data.

10. A micro-magnetic and three-dimensional point cloud fusion-based scraper conveyor damage identification system, characterized in that, The method for damage identification of scraper conveyors based on the fusion of micromagnetism and three-dimensional point cloud as described in any one of claims 1-9 includes: The sensing and acquisition module is used to acquire magnetic signal data and three-dimensional point cloud data of key components of the scraper conveyor; The synchronous registration module is used to preprocess data to generate magnetic space anchor points and geometric space anchor points, and to calculate and establish the correlation mapping matrix; The fusion analysis module is used to generate surface depression depth parameters, perform inverse threshold compensation to generate corrected high-frequency signals, extract and splice them to generate joint feature vectors, and call a pre-set damage classification model to generate damage classification results. The intelligent output module is used to generate visual damage location tag data.

Images